less than 1 minute read

Índice

Nuevo dataset

Ampliación de piloto manual

Remote pilot distribution

Pesos:

ANGULAR_UMBRALS = [-0.7, -0.2, 0, 0.2, 0.7, float('inf')] 
LINEAR_UMBRALS = [3.0, 4.25, float('inf')]

weights = [(0.2, 0.15, 0.0),
           (0.95, 0.55, 0.45),
           (0.75, 0.95, 0.75), 
           (0.75, 0.95, 0.75), 
           (0.95, 0.55, 0.45), 
           (0.2, 0.15, 0.0)] 

Dataset con piloto experto

Expert pilot distribution

Pesos:

MAX_ANGULAR = 0.5
MAX_LINEAR = 5.5
MIN_LINEAR = 1.0

ANGULAR_UMBRALS = [-0.45, -0.15, 0, 0.15, 0.45, float('inf')]
LINEAR_UMBRALS = [2.0, 3.25, float('inf')]

weights = [(0.2, 0.1, 0.0),
           (0.55, 0.65, 0.25), 
           (0.95, 0.95, 0.75), 
           (0.95, 0.95, 0.75), 
           (0.55, 0.65, 0.25),  
           (0.2, 0.1, 0.0)] 

Dataset combinado

Combined distribution
MAX_ANGULAR = 1.5
MAX_LINEAR = 6.0
MIN_LINEAR = 1.0

ANGULAR_UMBRALS = [-0.45, -0.15, 0, 0.15, 0.45, float('inf')]
LINEAR_UMBRALS = [2.0, 3.25, float('inf')]

weights = [(0.1, 0.2, 0.3), 
           (0.55, 0.75, 0.45),  
           (0.95, 0.85, 0.75), 
           (0.95, 0.85, 0.75), 
           (0.55, 0.65, 0.45), 
           (0.1, 0.2, 0.3)] 


Pruebas redes neuronales

A pesar de una mejora más definida en la augmentation, los resultados fueron bastante desfavorables en las 3 pruebas. Se cree que esto se debe a que se requerirá una red neuronal más compleja como DeepPilot para un mejor rendimiento.